Managing for late-successional/old-growth characteristics in northern hardwood-conifer forests William S. Keeton *

Forest Ecology and Management 235 (2006) 129–142
www.elsevier.com/locate/foreco
Managing for late-successional/old-growth characteristics
in northern hardwood-conifer forests
William S. Keeton *
Rubenstein School of Environment and Natural Resources, University of Vermont, Burlington, VT 05405, United States
Received 19 March 2006; received in revised form 2 August 2006; accepted 2 August 2006
Abstract
In the northern hardwood region of North America managing for late-successional forest habitats and functions is an important element of
ecosystem management. This study tests the hypothesis that uneven-aged practices can be modified to accelerate rates of late-successional forest
development. An approach, termed ‘‘structural complexity enhancement’’ (SCE), is compared against conventional uneven-aged systems modified
to increase post-harvest structural retention. Experimental treatments, including controls, were applied to 2 ha units and replicated at two multiaged northern hardwood forests in Vermont, USA.
Structural objectives include vertically differentiated canopies, elevated large snag and downed log densities, variable horizontal density
(including small gaps), and re-allocation of basal area to larger diameter classes. The latter objective is achieved, in part, by cutting to a rotated
sigmoid diameter distribution. This is generated from a basal area (34 m2 ha1) and tree size (90 cm dbh) indicative of old-growth structure. Forest
structure data have been collected over 2 years pre-treatment and 3 years post-treatment. Fifty-year simulations of stand development were run in
NE-TWIGS and FVS comparing treatment and no treatment scenarios. Simulations also tested the sensitivity of large tree development to
prescription parameters. Leaf area index retention was spatially variable but significantly (P < 0.001) greater under SCE (91%) compared to
conventional treatments (75%). Post-harvest aboveground biomass (P = 0.041), total basal area (P = 0.010), and stem density (P = 0.025) were
significantly different among treatments, with SCE generally retaining more structure than conventional treatments. SCE increased coarse woody
debris volumes by 140%; there was a 30% increase under conventional treatments. SCE successfully achieved the rotated sigmoid diameter
distributions, and sustained these 50 years into the future, resulting in reallocated basal area. Cumulative basal area increments are projected to
increase by 3.7 and 5.0 m2 ha1 compared to no treatment scenarios for SCE and conventional treatments, respectively. Basal areas will be
significantly (P = 0.025) greater after 50 years in SCE units due to higher residual basal areas. Conventional treatments are projected to produce 10
fewer large trees per hectare (>50 cm dbh) than would have developed without treatment, whereas SCE is likely to recruit five more large trees per
hectare than the no treatment scenario. Large tree recruitment rates were related primarily to the form of residual diameter distributions (P = 0.006)
and, possibly, to maximum diameter limits. Late-successional characteristics in northern-hardwood systems can be promoted through a variety of
modified uneven-aged silvicultural approaches based on the results.
# 2006 Elsevier B.V. All rights reserved.
Keywords: Old-growth; Late-successional; Northern-hardwoods; Uneven-aged silviculture; Forest structure and development; Sustainable forestry
1. Introduction
Sustainable forestry practices across managed forest landscapes contribute to the maintenance of biological diversity and
ecosystem functioning (Lindenmayer and Franklin, 2002). The
challenge lies in determining the mix of management
approaches – including type, timing, intensity, and spatial
configuration of silvicultural treatments – necessary to achieve
sustainability objectives. One possibility is to focus on the
* Tel.: +1 802 656 2518; fax: +1 802 656 2623.
E-mail address: [email protected].
0378-1127/$ – see front matter # 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.foreco.2006.08.005
architecture of individual forest stands and their spatial
arrangement, with consideration given to the aggregate
representation of multiple structural (or habitat) conditions at
landscape scales. Patch and successional dynamics associated
with natural disturbance regimes provide a useful guide for
designing this type of structure (Keeton, 2005) or disturbancebased (Mitchell et al., 2002; Seymour et al., 2002) approach. A
recommendation is to manage for currently under-represented
structures and age classes on some portion of the landscape
(Franklin et al., 2002; Keeton, 2005). In the northern hardwood
region of the eastern U.S. and Canada, this would include
managing for late-successional structure, which is vastly underrepresented relative to pre-European settlement conditions
130
W.S. Keeton / Forest Ecology and Management 235 (2006) 129–142
(Mladenoff and Pastor, 1993; Cogbill, 2000; Lorimer, 2001;
Lorimer and White, 2003). In this study variants of unevenaged silviculture are explored as a means for promoting latesuccessional/old-growth structural characteristics in northern
hardwood-conifer forests. Alternate approaches for accelerating stand development rates are tested.
1.1. Why manage for old-growth structure in northern
hardwood forests?
In the northeastern United States there has been considerable
debate regarding differing proposals for adjusting age class
distributions on forested landscapes. Forest structure and
composition in pre-European settlement landscapes were
spatially and temporally variable due to geophysical heterogeneity, climate variability, and disturbances, both natural and
anthropogenic (Foster and Aber, 2004). However, the relative
availability of stand structures has changed as a result of 19th
century forest clearing, agricultural land-use, land abandonment, subsequent reforestation, and 20th century forest
management. In northern hardwood forests, age-class distributions have shifted from a predominance, pre-settlement, of oldgrowth forests to a present dominance of second growth, 40–
80-year old forests (Lorimer and Frelich, 1994; Lorimer, 2001;
Lorimer and White, 2003). Lorimer and White (2003) estimate
that 70–89% of pre-settlement northern hardwood forests were
old-growth (uneven-aged, >150 years old), whereas young
forests (up to 15 years old) comprised 1–3% of those systems.
Definitions of old-growth northern hardwood-conifer forests
generally use a combination of age (ca. >150 years), human
disturbance history, and structure (Hunter, 1989; Dunwiddie
et al., 1996; Hunter and White, 1997). Today forests with these
characteristics occupy less than 0.5% of the region (Davis,
1996). Riparian functions (Keeton et al., in review), habitat
values (see reviews in Tyrrell and Crow, 1994b; Keddy and
Drummond, 1996; McGee et al., 1999), and carbon sequestration (Harmon et al., 1990; Krankina and Harmon, 1994; Turner
and Koerper, 1995; Strong, 1997; Houghton et al., 1999)
associated with late-successional forests have declined as a
result.
With natural reforestation and successional development on
old-fields, the availability of grassland and early-successional
forested habitats declined significantly during the 20th century.
Early-successional habitats are now re-approaching (Lorimer,
2001) or, in some locales, possibly even below (DeGaaf and
Yamasaki, 2003) pre-settlement levels. Patch-cut or largegroup selection harvesting methods are sometimes advocated to
enhance the relative abundance of early-successional habitats
(Hunter et al., 2001; King et al., 2001; Litvaitis, 2001; DeGaaf
and Yamasaki, 2003). Similarly, researchers have suggested
that a portion of the landscape could be managed for latesuccessional and old-growth forests (Singer and Lorimer, 1997;
Lorimer and White, 2003; Keeton, 2005). These proposals need
not be mutually exclusive, but do require differing silvicultural
approaches where active manipulations are desired. New or
modified approaches are needed to manage for late-stand
development characteristics, however, because conventional
even- and uneven-aged systems provide only limited availability of these structures (Gore and Patterson, 1985; McGee
et al., 1999; Crow et al., 2002; Angers et al., 2005).
1.2. Structural complexity enhancement
Despite theoretical discussions (e.g. Lorimer and Frelich,
1994; Trombulak, 1996), old-growth forest restoration techniques have not been experimentally field-tested in northern
hardwood forests. Thus, it remains uncertain whether silvicultural practices can accelerate rates and processes of latesuccessional forest stand development (Franklin et al., 2002),
promote desired structural characteristics, and enhance associated ecosystem functions more than conventional systems. A
related, though untested, hypothesis is that active restoration
offers advantages over passive (or non-manipulative) restoration
as means for recovering old-growth forest conditions. For
restorative silviculture to have more than only narrow appeal, it
must also provide opportunities for low-intensity timber harvest.
This study tests the ability of a variant of uneven-aged systems,
termed structural complexity enhancement (SCE), to achieve
these objectives (Table 1). In this study, SCE is compared against
conventional uneven-aged systems (Leak et al., 1987) modified
Table 1
Structural objectives and the corresponding silvicultural techniques used to promote those attributes in structural complexity enhancement
Structural objective
Silvicultural technique
Vertically differentiated canopy
Single tree selection using a target diameter distribution
Release advanced regeneration
Regenerate new cohort
Elevated large snag densities
Elevated downed woody debris densities and volume
Girdling of selected medium to large sized, low vigor trees
Felling and leaving trees, or
Pulling over and leaving trees
Variable horizontal density, including small canopy gaps
Harvest trees clustered around ‘‘release trees’’
Variable density marking
Re-allocation of basal area to larger diameter classes
Rotated sigmoid diameter distribution
High target basal area
Maximum target tree size set at 90 cm dbh
Accelerated growth in largest trees
Full and partial crown release of largest, healthiest trees
W.S. Keeton / Forest Ecology and Management 235 (2006) 129–142
to increase post-harvest structural retention and to represent best
management practices. Group-selection treatments are modified
to approximate the average canopy opening size associated with
fine-scale natural disturbance events in New England, based on
the findings of Seymour et al. (2002).Research over almost three
decades has described the characteristics and dynamics of oldgrowth northern hardwood and mixed northern hardwoodconifer forests across a range of geographic settings and
disturbance histories (e.g. Whitney, 1984; Gore and Patterson,
1985; Foster, 1988; Hunter, 1989; Woods and Cogbill, 1994;
Tyrrell and Crow, 1994a,b; Dahir and Lorimer, 1996; Hunter and
White, 1997; Goodburn and Lorimer, 1998, 1999; McGee et al.,
1999; Hale et al., 1999; McLachlan et al., 2000; Ziegler, 2000;
Crow et al., 2002; Angers et al., 2005; Keeton et al., in review).
Structural objectives for SCE are derived from this body of
research. They include vertically differentiated canopies,
elevated large snag and downed log volumes and densities,
variable horizontal density (including canopy gaps), and reallocation of basal area to larger diameter classes (Table 1). The
latter objective is achieved, in part, using an unconventional
guiding curve based on a rotated sigmoid target diameter
distribution.
Rotated sigmoid diameter distributions have been widely
discussed in the theoretical literature (e.g. Goff and West, 1975;
Nyland, 1998; O’Hara, 1998; Leak, 2002) but their silvicultural
utility has not been field tested. Sigmoidal form is one of several
possible distributions in eastern old-growth forests (Meyer and
Stevenson, 1943; Lorimer, 1980; Goodburn and Lorimer, 1999).
These vary with disturbance history, species composition, and
competitive dynamics (Goff and West, 1975; Leak, 1996, 2002).
It is uncertain whether sigmoidal diameter distributions can be
sustained silviculturally versus reverting to a negative exponential form. The latter would be expected if density-dependent
mortality (or self-thinning) proves constant across all size classes
(O’Hara, 1998). This study tests the hypothesis that the rotated
sigmoid distribution offers advantages for late-successional
structural management because it allocates more growing space
and basal area to larger size classes. I predict this distribution is
sustainable in terms of recruitment, growth, and mortality. If so, it
would support O’Hara’s (1998, 2001) assertion that there are
naturally occurring alternatives to the negative exponential or
‘‘reverse-J’’ curve used in uneven-aged silviculture.
2. Methods
2.1. Experimental design
The study was conducted at the Mount Mansfield State
Forest (MMSF, 448300 23.0300 N; 728500 11.2400 W) and at the
University of Vermont’s Jericho Research Forest (JRF,
448260 43.7000 N; 728590 44.1500 W). The former is located on
the western slopes and the latter resides in the foothills of the
northern Green Mountain Range in Vermont, USA. Elevations
range from 470 to 660 m (MMSF) and from 200 to 250 m (JRF)
above sea level. Soils are primarily Peru extremely stony loams
(MMSF) and Adams and Windsor loamy sands or sandy loams
(JRF). Study areas are northern hardwood-conifer forests;
131
dominant canopy trees are approximately 70–100 years old.
Tree demography is distinctly multi-aged due to regeneration
resulting from four to six documented management entries
since the early 20th century. Multi-aged structure was
confirmed by extensive pre-treatment tree coring across size
classes. Dominant overstory species include Acer saccharum
(sugar maple), Fagus grandifolia (American beech), and Betula
alleghaniensis (yellow birch). Tsuga canadensis (eastern
hemlock) is also co-dominant at JRF. There are minor
components of Picea rubens (red spruce) at MMSF and Acer
rubrum (red maple) and Quercus rubra (red oak) at JRF.
There were three experimental manipulations randomly
assigned to treatment units. Treatment units were 2 ha in size
and separated by 50 m (minimum) unlogged buffers to
minimize cross contamination of treatment effects. The first
two manipulations were conventional uneven-aged systems
(single-tree selection and group-selection) modified to increase
post-harvest structural retention. The modifications were based
on a target residual basal area of 18.4 m2 ha1, maximum
diameter of 60 cm, and q-factor (the ratio of the number of trees
in each successively larger size class) of 1.3. The group
selection treatment was based on the same BDq prescription but
applied through spatially aggregated harvesting. Individual
groups were placed to: (a) be well-distributed, (b) emcompass
the range of tree diameters needed to achieve the prescription,
and (c) release advanced regeneration. Group-selection cutting
patches averaged approximately 0.05 ha in size which resulted
in eight to nine groups per treatment unit. Slash and
unmerchantable upper tree boles were retained by the
conventional treatments, but there were no additional requirements for coarse woody debris (CWD) retention.
The third treatment was SCE. The guiding curve, based on a
rotated sigmoid target diameter distribution, was applied as a
non-constant q-factor: 2.0 in the smallest sizes classes, 1.1 for
medium-sized trees, and 1.3 in the largest size classes. The
guiding curve was also derived from a target basal area
(34 m2 ha1) and maximum diameter at breast height (90 cm)
indicative of old-growth structure. As a target, these parameters
define the form of the desired future diameter distribution.
Superimposing the target on pre-harvest diameter distributions
results in cutting to a residual basal area below both the target
and pre-harvest basal area. This is because target large tree
classes are unoccupied (i.e. pre- and immediately post-harvest),
pending future large tree recruitment. However, the prescription also resulted in retention of all trees >60 cm dbh. Full
(three- or four-sided) and partial (two-sided) crown release
were employed to accelerate growth in larger trees. To generate
site-appropriate CWD enhancement prescriptions, the differences between literature-derived targets and pre-harvest
volumes and densities were determined by unit. Snags were
created by girdling diseased, dying, or poorly formed trees. Pretreatment densities of low vigor trees were sufficient such that
girdling of healthy trees was not necessary to achieve snag
prescriptions. On one SCE unit at each of the two study areas,
downed logs were created by pulling (skidder and cable) or
pushing (mechanized tree shear) trees over, rather than felling,
to create pits and exposed root wads. Natural tip-ups were
132
W.S. Keeton / Forest Ecology and Management 235 (2006) 129–142
observed at these sites and deemed likely to occur due to
shallow bedrock and/or moderately to very mesic soils.
Each of the first two treatments (uneven-aged) was replicated
twice at MMSF; the third (SCE) was replicated four times, twice
at each of the two study areas. Two un-manipulated control units
were located at each study area. Experimental manipulations (i.e.
logging) were conducted on frozen ground in winter (January–
February) of 2003. All treatments included retention of
American beech exhibiting resistance to beech bark disease
(Nectria coccinea). Scattered mature red spruce were retained as
seed trees to encourage recolonization where this species was
historically more abundant.
2.2. Data collection
Each treatment unit contains five, randomly placed, 0.1 ha
permanent sampling plots. These are positioned at least 15 m
inside of unit boundaries, and collectively represent 25% of each
unit’s area. Within each plot, all live and dead trees >5 cm dbh
and >1.37 m tall were permanently tagged, measured, and
recorded by species, diameter, height, and vigor/decay stage (1–
7). Tree heights were measured using an Impulse 200 laser range
finder. Canopy closure was measured with a spherical
densitometer at 13 systematically placed points nested within
each overstory plot (n = 65 per unit). Plot and tree positions were
geo-referenced using a Trimble Pro XRS Global Positioning
System. We also used the GPS to map the perimeter of group
selection cutting patches. Downed log (logs > 10 cm diameter)
volume by decay class (1–5) was estimated using a line-intercept
method (two 31.62 m transects per plot) following Shivers and
Borders (1996). Log densities were inventoried across 0.1 ha
plots. Snag and downed log decay classes followed Sollins et al.
(1987). Leaf area index (LAI) at ground level was measured at
five points in each plot (n = 25 per unit) using a Li-Cor 2000 Plant
Canopy Analyzer. LAI values were post-processed to calibrate
‘‘below canopy’’ measurements against ambient ‘‘above
canopy’’ light measurements taken by a remotely placed LiCor meter. Tip-up mounds were inventoried across each unit and
measured in three dimensions. Two dominant canopy trees per
plot were cored at breast height to allow subsequent laboratory
determination of tree age. Two years of pre-treatment and 3 years
of post-treatment data collection have been completed. Two
sample plots (one SCE unit and one single-tree selection unit)
escaped treatment due to inoperable or wet terrain. Data from
these plots were not included in analyses of post-treatment data.
During treatment implementation, harvested logs were
sorted by product grade or type and treatment unit at the
landing. Logs were then transported and tracked independently
by unit through to scaling at the processing mill. In this way
harvest volumes (based on mill receipts rather than inventory
data) could be determined by treatment. Volumes reported here
are based on a conversion factor of 4.53 m3 per 1000 board feet.
2.3. Data analysis
Pre- and post-harvest sample data were input into the
Northeast Decision Model (NED-2) (Twery et al., 2005), which
was used to generate a suite of stand structure metrics. These
included aboveground biomass estimates based on speciesspecific allometric equations developed by Jenkins et al.
(2003). Structural metrics were compared pre- to post-harvest
and among treatments using a before/after/control/impact
design (Krebs, 1999). Tukey-tests were used for the former
while analysis of variance and post hoc Bonferroni (CWD data)
or least significant difference (LSD, overstory data) multiple
comparisons were used for the latter. Homogeneity of variance
was tested using F-tests for both pre- and post-harvest data.
‘‘Site’’ was not modeled as a random effect due to low sample
size and incomplete replication of treatments across sites. The
error term in one-way ANOVAs was thus mean square error.
Multiple comparisons were used to validate the assumption of
consistency in structural parameters among similarly treated
units, including controls, at different sites. In addition, spatial
autocorrelation tests (Ripley, 1981) using the Moran coefficient
were performed on key response variables. Response data were
sorted by treatment and made spatially explicit using georeferenced plot positions. Spatial autocorrelation results were
cross-checked against empirical variograms produced using SPlus software. Pre- to post-harvest shifts in downed woody
debris decay class distributions were assessed using the
Kolmogorov–Smirnov two-sample goodness of fit test.
A focused assessment of diameter distribution (5-cm diameter
classes) responses was conducted to determine whether SCE
successfully shifted diameter distributions, immediately postharvest, towards the target rotated sigmoid form. Pre- and postharvest and target distributions were log transformed to enhance
sigmoidal tendencies (Leak, 2002). Residual distributions were
smoothed using a Friedman smoothing run in S-Plus software.
Differences between transformed residual and target cumulative
frequency distributions were assessed using Kolmogorov–
Smirnov two-sample goodness of fit tests. Residual distributions
were created using real (sample) data for smaller diameter
classes (<70 cm dbh) and hypothetical (e.g. future potential)
values for larger diameter classes (>70 cm dbh). The latter
borrowed values from the target distribution. Statistical tests
evaluated whether residual distributions achieved that portion of
the target distribution possible given the pre-harvest structure.
NED-2 output was used for simulation modeling of stand
development using two models: the northeastern U.S. variant
(NE-FVS; Bush, 1995) of the USDA Forest Service’s Forest
Vegetation Simulator (Dixon, 2003) and NE-TWIGS (Hilt and
Teck, 1989). The NE-FVS modeling structure is based on NETWIGS, which is an individual tree-based, distance-independent
stand growth simulator. The models are empirical, with
coefficients fitted from repeated measurements in permanent
sampling plots encompassing more than 90,000 trees (Bankowski
et al., 1996). Invalidation tests, NE-TWIGS has proven somewhat
more reliable at estimating growth in uneven-aged stands
compared to even-aged stands (Bankowski et al., 1996). Mortality
and large-tree growth functions operate slightly differently in NEFVS and calculations are made every 10 years, rather on the
annual time step employed in NE-TWIGS. Site index50 (a
required input parameter) was held constant and set to 19.8 m for
sugar maple, representing a moderately productive site.
W.S. Keeton / Forest Ecology and Management 235 (2006) 129–142
Fifty-year projections of stand development were run for each
treatment unit, including controls. To compare relative
restorative potential, only stand development resulting from
initial treatment (rather than periodic re-cuttings) was modeled
(see Section 4.3). For manipulated units, both no treatment
(based on pre-harvest sample data) and treatment (based on postharvest sample data) scenarios were simulated. Mean projected
diameter distributions by treatment were evaluated to determine
whether either rotated sigmoid (SCE) or negative exponential
(conventional uneven-aged) were sustained over time. Projected
basal area distributions were generated to determine the
corresponding effect on basal area re-allocation. To evaluate
projected growth responses, I calculated cumulative basal area
increment (CBAI) for each simulation run at 5-year intervals. To
normalize treatment scenarios against site/unit-specific rates of
stand development that could be expected without treatment, I
calculated differences in CBAI between ‘‘no-treatment’’ and
‘‘treatment’’ scenarios at each time step. The Kolmogorov–
Smirnov two-sample goodness of fit test was used to test for
differences between treatment groups along mean CBAI time
series. The log-likelihood ratio goodness of fit test (G test) was
used to examine total CBAI developed after 50 years; response
ratios (treatment versus no treatment) were compared against a
null ratio (no treatment effect). Simulation modeling also
generated predictions regarding the number of large trees (two
classes: >50 and >60 cm dbh) that might develop after 50 years.
Additional simulations were run to determine the sensitivity
of projected large tree densities and basal area allocations to
prescriptive diameter distributions and maximum diameter
limits. These simulated stand development for each of the four
SCE units when cut to: (1) a negative exponential (qfactor = 1.3) distribution with no maximum diameter limit
(i.e. set to 90 cm to generate the distribution), (2) a negative
exponential distribution with a maximum diameter of 60 cm,
and (3) a rotated sigmoid distribution with a maximum
diameter of 60 cm. When combined with the actual SCE
treatment (rotated sigmoid, no maximum diameter limit), this
resulted in a factorial design with two levels of each
independent variable. Simulated cuts were conducted by
deleting (random selection by tree tag #) or adding (random
selection of harvested trees by tag #) the number of trees
required (by diameter class) to change post-harvest diameter
distributions to the desired form. Additions and deletions were
evenly distributed across plots. Where pre-harvest distributions
were in deficit for particular size classes, as with actual
treatment, post-harvest densities were not modified. Thus, the
sensitivity analysis evaluated alternate post-harvest distributions that could be created given real pre-treatment structure.
Residual basal area was held constant. Simulation results were
evaluated using two-way Analysis of Variance.
Single-tree selection and group selection treatments were
classified as one group (‘‘conventional uneven-aged’’) for
analyses of simulation output (e.g. calculations of mean
projections by treatment) and other statistical tests. This was
appropriate because: (1) there were no significant differences
among uneven-aged units in residual structure when data were
aggregated to the unit scale; and (2) quantitative prescriptions
133
were the same for both, the only difference being dispersed
versus aggregated harvesting. This difference was not relevant
since neither NED-2 output nor the simulations were spatially
explicit. Spatial attributes of group selection cutting patches
and correspondence with sampling plots were analyzed in a
geographic information system.
3. Results
3.1. Pre-harvest structure
Pre-treatment structure was not significantly different
(a = 0.05) among treatment units based on ANOVA results.
This held for basal area, aboveground biomass, stem density,
and all other variables tested (Table 2). Variance in structure
among units assigned to the same treatment was not statistically
significant (a = 0.05). These results suggest there is a
statistically valid basis for comparisons of both pre- to postharvest changes and with respect to differences in residual
structure among treatment units.
3.2. Timber harvest volume
Timber harvest volumes varied by treatment and unit.
Single-tree selection averaged 34.7 (minimum = 30.5, maximum = 39.0) m3 ha1 in volume harvested. The average
volume harvested from group selection units was 20.0
(minimum = 19.9, maximum = 20.2) m3 ha1, and averaged
Table 2
Pre- to post-treatment changes in response data for key structural response
variables
Controls
Mean
SCE
S.E.
a
Aboveground biomassb (kg ha1)
Pre-harvest
10026.2 1016.1
Post-harvest
10165.4 1087.1
Percent change
1.4
Conventional
Mean
S.E.a
985.4
808.7
11163.6
6398.1
42.7
570.8
691.7
Mean
S.E.
11313.0
8646.3
23.6
a
Basal areab (m2 ha1)
Pre-harvest
31.4
Post-harvest
31.4
Percent change
0.2
2.0
1.8
35.6
26.5
25.6
3.2
3.2
32.2
18.6
42.3
1.0
1.6
Canopy cover (%)
Pre-harvest
Post-harvest
Percent change
4
4
95
77
19
2
3
96
69
28
2
2
98.7
84.6
1044.0
750.0
28.2
4.5
41.2
958.5
564.1
41.1
107.0
95.6
5.3
4.2
41.2
91.6
140.0
7.7
11.4
38.2
49.0
30.0
3.1
2.2
92
92
0
Stem densityb (trees ha1)
Pre-harvest
982.0
Post-harvest
934.5
Percent change
4.8
Downed log volume (m3 ha1)
Pre-harvest
42.0
Post-harvest
43.1
Percent change
4.0
Means and errors are assessed among treatment units (n = 4).
a
1S.E. of the mean.
b
Live and dead trees >5 cm dbh.
134
W.S. Keeton / Forest Ecology and Management 235 (2006) 129–142
16.4 (minimum = 11.6, maximum = 20.7) m3 ha1in SCE
units. Variation among units sharing similar treatment was
due, primarily, to differences in timber quality (e.g. stem defect
and form); this determined allocation to sawlog and veneer
grade (i.e. board foot) production versus cordwood (i.e.
firewood) production. Harvest volumes for SCE do not include
trees allocated to CWD, which were included in the total
number of stems cut to achieve the target diameter distribution.
SCE produced 53% of the average saw/veneer log volume
harvested by conventional uneven-aged treatments.
Normalizing the data to reflect differences in pre-treatment
(or available) volume yields a slightly different perspective. In
SCE units 19% of the pre-harvest merchantable volume
(85.8 m3 ha1) was removed on average. Considering preharvest merchantable volume in single-tree (96.9 m3 ha1) and
group selection (62.9 m3 ha1) units, 36 and 32% of the
available volume was cut, respectively. From this perspective,
SCE resulted in a harvesting intensity about 57% of that
incurred by the conventional treatments.
3.3. Residual stand structure
All of the experimental treatments maintained high levels of
residual structure, although there were distinct differences.
There were no significant differences (a = 0.05) in post-harvest
structure between similarly treated units based on F-tests of
variance. There was no significant (a = 0.05) spatial autocorrelation among plots sorted by treatment. This result was
confirmed by empirical variograms. These results suggest that
similar treatments produced consistent effects with respect to
horizontal structure and spatial heterogeneity. In group
selection units on average, 30% of the area in cutting patches
coincided with sampling plots and 32% of total plot area was in
cut patches. Of the total area in these units 35% was cut on
average. Thus cut areas were slightly under-sampled and postharvest structure was slightly over-estimated for the group
selection treatment.
There were significant differences (P < 0.001) in treatment
effects on leaf area index (LAI). LAI values decreased pre- to
post-harvest from 6.27 to 5.05 (single-tree selection) and from
6.49 to 4.79 (group selection). This represented reductions of
20 and 30% respectively. LAI reductions were lowest in SCE
units (9%), falling from 5.75 to 5.15. Changes (pre- to postharvest) in LAI were significantly more spatially variable for
both SCE (P = 0.031) and group-selection (P = 0.010) compared to single-tree selection. Variability in LAI change was not
significantly different between SCE and group-selection units
(P = 0.296). These results reflect the high degree of horizontal
structural variability associated with aggregated harvesting in
group-selection. In SCE units, smaller (<0.05 ha on average)
gaps and canopy openings were created through variable
density marking and clustered harvesting around crown-release
trees. LAI in control units increased from 5.33 to 5.35 over the
course of this study; this change was not statistically significant.
Both SCE and conventional uneven-aged treatments resulted
in structural changes pre- to post-harvest (Table 2), but these were
statistically significant only for the conventional treatments
(P < 0.001). Stand structure did not change significantly in
control units during the course of this study. Post-harvest
aboveground biomass (P = 0.041), total basal area (P = 0.010),
and stem density (P = 0.025) were significantly different among
treatments based on ANOVA and post hoc comparisons.
Conventional treatments resulted in significantly lower aboveground biomass (P = 0.014), total basal area (P = 0.003), and
stem density (P = 0.008) in comparison to control units. SCE did
not result in statistically significant contrasts with controls.
Residual basal area (P = 0.037) was significantly greater in SCE
units compared to conventional uneven-aged units; post-harvest
biomass (P = 0.104) and stem density (P = 0.124) were not
significantly different between these treatments.
There were no statistically significant differences (P > 0.05)
between residual and target distributions for any of the SCE
units based on comparisons of log transformed diameter
distributions (Fig. 1). This suggests that SCE was successful at
shifting residual diameter distributions towards the target
rotated sigmoid form. The smallest trees (5–10 cm dbh) were
undercut by all the treatments due to normal operational
limitations (see Nyland, 1998).
3.4. Canopy structure
Post-harvest canopy closure was significantly (P < 0.01)
greater for SCE (mean 77%) and group-selection (mean 73%)
compared to single-tree selection (mean 64%); SCE and group
selection were not significantly different in this respect.
Aggregate canopy closure remained relatively high following
group-selection because 70–80% of each unit retained full preharvest structure. However, spatial variability in canopy closure
was also greatest across group-selection units. In SCE units,
variable density marking achieved a similar though less
pronounced effect.
3.5. Crown release
Clustered harvesting in SCE units resulted in crown release
around 45 dominant and co-dominant trees per hectare on
average. The average pre-treatment number of large trees (not
released) was 20 ha1, so our future target of 55 large
(>50 cm dbh) trees per hectare was exceeded when these
densities are combined. The excess provides a ‘‘margin of
safety’’ to accommodate canopy mortality. Canopy trees were
released across a range of diameter classes (>25 cm dbh); the
majority (79%) were fully, rather than partially, crown released.
3.6. Coarse woody debris enhancement
The treatments had distinctly different effects on CWD
availability, both in terms of standing dead trees and downed
logs. SCE prescriptions increased CWD densities by 10 boles
(>30 cm dbh) per hectare on average for snags and 12 boles
(>30 cm dbh) per hectare on average for downed logs. Post
harvest dead tree basal area was 39% greater in SCE compared
to conventional treatments, although this was not statistically
significant (P = 0.092) based on multiple comparison results.
W.S. Keeton / Forest Ecology and Management 235 (2006) 129–142
135
Fig. 1. Pre-harvest, post-harvest, and target diameter distributions (cm dbh) for the four SCE units, two at Mt. Mansfield (above) and two at the University of
Vermont’s Jericho Research Forest (below). Log transformed post-harvest and target distributions are compared (top portion of graphs) using the Kolmogorov–
Smirnov two-sample goodness of fit test. There were no statistically significant differences. Thus, the post-harvest distributions achieved the target rotated sigmoid
distribution.
Post-harvest dead basal area was significantly greater when
SCE units were compared to controls (P = 0.036); conventional
units showed no significant differences with controls.
There were statistically significant differences (P = 0.002)
among treatments with respect to effects on downed log volume
(Fig. 2). SCE increased downed log volumes by 140% on
average. Volumes increased from 41 m3 ha1 pre-harvest to
92 m3 ha1 post-harvest. Mean downed log volume increased
30% in the uneven-aged units, changing from 38 m3 ha1 preharvest to 49 m3 ha1 post-harvest, although this effect was not
statistically significant relative to the controls. There were
slight (4% mean) increases in two of the four control units
caused by windthrow. Background recruitment rates were thus
not sufficient to explain SCE treatment effects. Analyses of
downed log decay class distributions in SCE units showed, as
expected, significant (P < 0.05) shifts towards less decayed
logs due large inputs of felled trees (Fig. 3).
Of 48 attempts (24 per study area) at pulling (Mt. Mansfield)
or pushing (Jericho) trees over, 45 were successful at creating
large (mean of 14 m3) exposed root wads and pits (mean of
approximation 7 m3). Success rates were influenced by the
relatively shallow depth to bed rock (0.5–1.2 m mean) at sites
where this was attempted. Tip-up mounds were significantly
larger ( p < 0.01) at Mount Mansfield, which was probably
related to the greater loam and moisture content of soils at
that site.
3.7. Projected stand development
Fig. 2. Downed log response to treatments. Shown are percent change from preharvest levels and absolute change in volume (m3 ha1). Error bars are 1S.E.
of the mean.
3.7.1. Basal area and aboveground biomass development
In comparison to NE-TWIGS, FVS (northeastern variant)
tended to produce more conservative estimates of growth
increment and large tree recruitment due to the differences
in model operation previously mentioned. Simulations in
136
W.S. Keeton / Forest Ecology and Management 235 (2006) 129–142
Fig. 3. Mean pre and post-treatment downed log (>10 cm) decay class distributions for SCE units. Cumulative frequency distributions were significantly
different based on Kolmogorov–Smirnov two-sample goodness of fit tests.
Decay class (DC) 1 is least decomposed while DC 5 is most decomposed. Note
the large input of fresh logs (DC 1) following treatment. Post-harvest increase in
DC 4 is related to recruitment from DC 3 and inputs of several large, welldecayed upper limbs from declining trees.
NE-TWIGS of stand development following SCE project, on
average, that basal areas will approach 34 m2 ha1 after 50
years (Fig. 4, top). This is 24% (or 8 m2 ha1) greater than the
mean for the conventional uneven-aged units. Projected basal
area for SCE also exceeds the mean predicted for control units
Fig. 4. Results of NE-TWIGS stand development modeling. Shown are 50-year
projections of post-treatment cumulative basal area (live tree) production (top)
and normalized scenarios (post-harvest minus pre-harvest cumulative basal area
increment, bottom). Error bars are 1S.E. of the mean.
by 13% (or 4.5 m2 ha1). Conventional units were projected to
have basal areas still 12% (or 3.6 m2 ha1) below the control
units after 50 years of development. Projected basal area was
more variable (Fig. 4, top) among SCE units (2.9 m2 ha1)
compared to conventional units (0.6 m2 ha1). Stand
development projections differed slightly depending on choice
of model. While the magnitude of difference among
treatments was consistent in FVS projections, CBAI and
end-of-run basal area were 12–18% lower, except for control
units in which CBAI was flat and thus consistent under both
models.
Projected tree growth rates did not differ statistically
between treatment scenarios, as measured by CBAI and
evaluated using goodness of fit tests (Dmax = 0.007, critical
value of D0.05 = 0.307). When projected development with
treatment is normalized, to reflect the amount of development
(specific to each unit) that would have been expected with no
treatment, the simulations indicate that CBAI will be slightly
faster under conventional systems (Fig. 4, bottom). However,
this difference was not statistically significant (Dmax = 0.005,
critical value of D0.05 = 0.183). Both SCE (P < 0.05) and
conventional treatments (P < 0.01) are projected to significantly accelerate tree growth rates above that expected with no
treatment based on both NE-TWIGS and FVS modeling.
Projected CBAI after 50 years was 2.7 m2 ha1 (no treatment)
compared to 6.4 m2 ha1 (with treatment), on average, in SCE
units. For conventional units projected CBAI increased from
0.34 m2 ha1 (no treatment) to 5.3 m2 ha1 (with treatment).
Therefore, differences in basal area development are largely
attributable to treatment effects on residual basal. Both
conventional and SCE treatments accelerate basal area
increment, but SCE leaves more residual basal area and thus
ultimately results in higher basal areas.
Neither SCE nor conventional treatments resulted in
projected basal areas that exceeded those projected for the
corresponding units under a ‘‘no treatment’’ scenario. However,
basal area in SCE units recovered to within 89% of the notreatment scenario, whereas conventional units recovered to
within 77% on average. This difference was statistically
significant (P = 0.025).
Aboveground biomass is predicted to increase over the
next 50 years in all of the experimental units, including
controls, based on the FVS simulations and NED-2 biomass
calculations. Living biomass increases 70.5, 66.1, and 73.6%
on average for SCE, uneven-aged, and control units
respectively under a ‘‘no treatment’’ scenario. It increases
90.7 and 105.2% following SCE and uneven-aged treatment,
but starts from a post-harvest level that is 18.3 and 35.9%
lower than pre-treatment respectively. Consequently, neither
treatments achieve the biomass they would have without
treatment. However, after 50 years SCE results in aboveground biomass that is 91.4% of that projected under no
treatment, while the conventional treatments result in 79.1%
of the no treatment potential. Biomass production annual
increment is accelerated 5.1% for SCE and 1.9% for
conventional treatments based on normalizing treatment
against no treatment scenarios.
W.S. Keeton / Forest Ecology and Management 235 (2006) 129–142
137
Table 3
Two-way ANOVA results exploring projection sensitivity to diameter distribution (rotated sigmoid vs. negative exponential) and use of maximum diameter limit
(60 cm dbh limit vs. no limit)
Factor
Difference from no treatment scenario for:
MS
F
P
Maximum diameter limit
Large tree density
Large tree basal area
29.675
6.428
0.770
1.826
0.398
0.202
Diameter distribution
Large tree density
Large tree basal area
436.706
36.552
11.327
10.384
0.006
0.007
Diameter distribution max
Diameter limit (interaction)
Large tree density
Large tree basal area
2.933
0.495
0.076
0.141
0.787
0.714
Results shown are for SCE units (n = 4) at both study areas. The tests evaluated the difference in large tree (>50 cm dbh) density and basal area projected 50 years
following treatment compared to an equal period of development with no treatment for the same experimental unit. Italicized values are statistically significant
(a = 0.05).
3.7.2. Large tree recruitment
Rates of large tree recruitment are likely to be faster under
SCE compared to no treatment scenarios. In SCE units, there
will be an average of five more large trees (>50 cm dbh) per
hectare than there would have been without treatment after 50
years, based on FVS projections. There will be 10 fewer large
trees per hectare on average in the conventional units than
would have developed in the absence of timber harvesting. SCE
results in an increase of four very large trees (>60 cm dbh) per
hectare, while conventional treatments produce three fewer
very large trees per hectare than would have been recruited
without treatment.
The sensitivity analysis showed no significant interaction
between residual diameter distribution and maximum diameter
limit (Table 3). Choice of target diameter distribution, however,
had a significant effect on large tree (>50 cm dbh) density
(P = 0.006) and basal area (P = 0.007) recruitment. Maximum
diameter limit also affected projected densities and basal areas
(Table 4), but these differences were not statistically significant
(Table 3). Increases in projected large tree densities (over that
projected for no treatment) after 50 years were highest when
both the rotated sigmoid and a very high (or no) maximum
diameter limit were employed (actual treatment). Adding a
diameter limit of 60 cm to the rotated sigmoid prescription
reduced large tree recruitment to three stems per hectare more
than would have developed with no treatment.
Simulated negative exponential distributions employed with
no diameter limit resulted in an intermediate level of large tree
recruitment, but nevertheless had fewer large trees than would
have developed with no treatment (Table 4). Large tree
densities were lowest for the combination of a negative
exponential distribution and a maximum diameter limit
(Table 4). This held true as an average and for three of the
four units evaluated. However, one unit (Mt Mansfield) showed
large tree recruitment under this scenario comparable to the
actual treatment due to: (1) a competitive response to the
removal of one very large tree, and (2) a general pre-harvest
deficit of trees >45 cm dbh. The latter limited the potential to
create substantial differences in post-harvest structures through
simulated treatments.
3.7.3. Projected diameter and basal area distributions
Projections suggest that a rotated sigmoid diameter
distribution will sustain itself over 50 years in SCE units
(Fig. 5). Projected and target (i.e. desired future) rotated
sigmoid distributions were not significantly different
(Dmax = 0.058, critical value of D0.05 = 0.430). This held as
long as densities in smallest two diameter classes (stems
<15 cm dbh) were held constant. This was appropriate because
both NE-TWIGS and FVS lack a seedling recruitment model
(there is only limited regeneration through stump spouting).
Pronounced regeneration deficiencies thus develop rapidly in
model projections. This explains the low densities evident in the
smallest size classes in Fig. 5. Uneven-aged units appear to
maintain the negative exponential distributions initially
prescribed. However, they have significantly less recruitment
into larger size classes compared to the projected distributions
for SCE (Fig. 5).
Differences in projected diameter distributions and large tree
recruitment explain related changes in basal area distributions.
Compared to stand development under a ‘‘no treatment’’
scenario, SCE results in significant reallocation of basal area
Table 4
Results of the sensitivity analysis exploring projected differences after 50 years of stand development between alternate treatment scenarios for SCE units (n = 4)
Treatment
Structural response
Residual diameter distribution
Use of maximum diameter limit
Large tree density (stems/ha)
Large tree basal area (m2/ha)
Rotated sigmoid
Rotated sigmoid
Negative exponential
Negative exponential
No
Yes
No
Yes
+5
+3
5
8
+2
+1
1
3
Values shown for large tree (>50 cm dbh) densities and basal areas are the mean change with treatment relative to development projected under no treatment (passive
management) for the same units.
138
W.S. Keeton / Forest Ecology and Management 235 (2006) 129–142
treatments – modified to include a low q-factor and relatively
high maximum diameter – also resulted in some reallocation of
basal to larger size classes (Fig. 6, bottom). However, this
reallocation was substantially lower than for SCE and did not
include recruitment into size classes larger than the projected
‘‘no treatment’’ distribution. In fact, the ‘‘no treatment’’
simulations showed basal area development into the 75–80 cm
diameter class, whereas this was lacking for the corresponding
units subjected to uneven-aged treatments (Fig. 6, bottom).
4. Discussion
Fig. 5. Projected diameter distributions (cm, dbh) for 50 years into the future
based on FVS simulations. Mean distributions developed after structural
complexity enhancement (SCE) and conventional uneven-aged treatments
are compared against the target or desired future condition prescribed for
SCE. Note the close fit with the target shown by SCE, and the disparity evident
for the conventional treatment.
into the largest size classes (e.g. >50 cm dbh). This includes a
shift of basal area into the very largest size classes
(>85 cm dbh) that experience no basal area recruitment under
the ‘‘no treatment’’ scenario (Fig. 6, top). The uneven-aged
Uneven-aged silvicultural techniques can be modified to
promote development of old-growth structural characteristics
in northern hardwood and mixed northern hardwood-conifer
forests. Both the uneven-aged and SCE approaches tested
maintain high post-harvest levels of some structural attributes,
such as stem density and canopy cover. However, SCE
maintains or supplements CWD volume, basal area, aboveground biomass, large tree recruitment, and other structural
attributes to a greater degree. Higher post-harvest LAI under
SCE signals greater retention of foliage bearing tree crowns,
representing an important element of vertical complexity
(Parker et al., 2004). In addition, SCE results in a rotated
sigmoid diameter distribution that appears self-maintaining at
least over 50 years, and consequently reallocates growing space
and aboveground structure into larger size classes. This
contributes to increased density of large trees, a higher foliage
biomass, and increased canopy complexity. The treatments are
likely to develop differently with respect to horizontal
complexity or patchiness. This inference is supported by the
observed spatial variability in LAI and canopy closure created
by group selection (most variable and contrasting) and SCE
(moderately variable). The finding of more spatially uniform
structure maintained by single-tree selection (least variable) is
supported by previous research (Kenefic and Nyland, 2000;
Crow et al., 2002).
4.1. Accelerating rates of stand development
Fig. 6. Projected live tree basal area distributions for 50 years into the future
based on FVS simulations of stand development. Mean distributions are shown
for SCE units: (A) and conventional uneven-aged units and (B) projected both
with and without treatment. No treatment scenarios are specific to treatment unit
(i.e. development that would have occurred without treatment in a given unit);
they are not based on data from control plots. Thus, the ‘‘no treatment’’
scenarios differ by treatment unit/type. Error bars are 1S.E. of the mean.
Accelerated tree growth can be expected after both SCE and
conventional uneven-aged treatments according to simulation
projections. Conventional treatments retain less basal area and,
thus, result in moderately greater projected basal area
increments. This finding is consistent with previous research
on growth responses to stocking density and growing space
availability (Leak et al., 1987). A shortcoming in the stand
development projections is that neither NE-TWIGS nor the
FVS model is spatially explicit. Individual tree growth rates
reflect competition only as a function of the total stand stocking
in trees of equivalent or greater diameter (Bush, 1995). The
models do not, therefore, capture the effects of crown release on
selected dominant trees as employed in SCE. Crown release has
been found to partially arrest or dampen declining growth rates
in older, dominant northern hardwoods, leading to rates of large
tree development that are 50–100% faster compared to no
release scenarios (Singer and Lorimer, 1997). Despite this
W.S. Keeton / Forest Ecology and Management 235 (2006) 129–142
limitation, the projections suggest that SCE does promote large
tree recruitment. Large tree recruitment will be significantly
impaired under conventional treatments. The results of the
sensitivity analysis suggest this is due to both maximum
diameter limits and negative exponential diameter distributions
that, collectively, provide lower large tree recruitment potential.
This holds even when the basal area prescription is the same as
under SCE.
Projected basal area and aboveground biomass are greater
after 50 years of development under SCE (in comparison to
conventional treatments) due, primarily, to elevated postharvest structural retention. Substantial increases in rates of
production over that likely without treatment also contribute to
these changes. However, while structural development in SCE
units is projected to surpass controls units, none of treatments
are likely to result in basal areas or aboveground biomass
exceeding levels that would have developed without treatment.
This provides a strong argument for passive restoration, rather
than silvicultural manipulation, as an approach for ultimately
developing greater levels of structural complexity associated
with these parameters. However, that conclusion does not
account for the accelerated rates of large tree recruitment,
reallocation of basal area, and associated structural complexity
projected for SCE. In this respect, active silvicultural
manipulation may offer some advantages.
It is also clear that the conventional uneven-aged approaches
were less effective at retaining and promoting some aspects of
late-successional/old-growth structural development. Periodic
harvests under a more standard entry cycle (e.g. 20 years)
would further limit basal area development and, if practiced
with a diameter limit, large tree density (Bryan, 2003). The
results do show that conventional approaches can be modified
to reallocate basal area to some degree and provide a relatively
high degree of biomass and vertical complexity. That singletree selection in northern hardwoods has the potential to
maintain vertical complexity is supported by previous research
(Kenefic and Nyland, 2000). However, uneven-aged practices
that employ maximum diameter limits impair and, in fact,
reduce large tree recruitment potential based on the results
reported here. In comparison, SCE ultimately results in greater
levels of structural complexity across a range of parameters.
Therefore, SCE offers a useful alternative for landowners
interested in both low-intensity harvesting and promotion of
old-growth characteristics.
4.2. Coarse woody debris dynamics and tip-up mounds
Coarse woody debris, both standing and downed, plays a
critical role in northern-hardwood systems as habitat and in
other ecosystem processes (Hunter, 1999). Techniques
employed in SCE were successful at creating the tip-up
mounds (i.e. root wads) common to late-successional northern
hardwood forests (Crow et al., 2002). Similar techniques have
been employed at the Harvard Forest, MA to simulate hurricane
effects (Carlton and Bazzaz, 1998), but this study is among the
first to experimentally manipulate tip-mound density as part of
commercial or restorative timber harvest. Tip-up mounds
139
provide unique microsite habitat characteristics and influence
soil nutrient processes (Beatty and Stone, 1986; Liechty et al.,
1992; Crow et al., 2002). It remains uncertain whether the
treatments will affect long-term mound recruitment dynamics.
SCE proved effective at enhancing CWD densities and
volumes above the immediately post-harvest levels associated
with slash and crowns left by the conventional treatments. Postharvest densities of downed logs and snags (>30 cm dbh),
including treatment additions, were substantially greater than
minimum levels suggested by some previous guidelines (e.g.
Elliott, 1988) but were consistent with more recent recommendations (e.g. TNC, 2001). Downed log volumes achieved
88% of the level recommended by McGee et al. (1999) to
approximate old-growth characteristics. They were about 50–
100% greater than volumes reported in mature northern
hardwoods managed under selection systems elsewhere
(Goodburn and Lorimer, 1998; Hale et al., 1999; McGee
et al., 1999).
It remains uncertain whether silviculturally elevated CWD
will persist until natural log recruitment rates increase, or,
alternatively, whether CWD enhancement in mature stands has
only transient or short-term management applications. As
silviculturally enhanced CWD decays it will become more
biologically available in habitat and nutrient processes (Gore
and Patterson, 1985; Tyrrell and Crow, 1994a; Goodburn and
Lorimer, 1998). Terrestrial salamander (McKenny et al., 2006)
and soil invertebrate (Donald R. Tobi, unpublished data)
populations are responding positively to increased downed log
densities in the treatment units.
Treatment effects on future snag and downed log recruitment are also uncertain, especially given the potential for
episodic recruitment associated with natural disturbance events
(Ziegler, 2002). If practiced with periodic harvest re-entries, all
of the treatments could result in subsequent CWD inputs and
maintenance of a range of decay classes (Kenefic and Nyland,
2000), although the net effect on CWD volume is likely to be
more limited under conventional systems (McGee et al., 1999).
In this study, silviculturally created snags and downed logs
were culled from the total number of trees harvested to achieve
target diameter distributions, and thus did not result in
additional removals of low-vigor trees beyond that associated
with conventional harvests. Nevertheless, preferential cutting
of low-vigor trees under any harvest system is likely to result in
a near-term reduction in CWD recruitment. Over the longerterm, however, increased densities of large trees associated with
SCE may also accelerate recruitment of large snags and logs.
4.3. Applications for structural complexity enhancement
SCE has a variety of useful applications, ranging from oldgrowth restoration, to riparian management, to low-intensity
timber management. These will depend greatly on economic
feasibility under a variety of operational scale, site quality,
product, and market conditions (Niese and Strong, 1992),
which is the subject of an on-going investigation (Keeton and
Troy, 2006). SCE could be employed to varying degrees
depending on the specific application. For instance, where
140
W.S. Keeton / Forest Ecology and Management 235 (2006) 129–142
timber production is emphasized, a subset of SCE elements
might be used. Other elements, such as CWD enhancement,
might be avoided or employed at a lesser intensity. In this
scenario, multiple stand entries would be expected, but latesuccessional structural development would be lower compared
to full SCE implementation. Such an approach, however, would
allow forest managers to build some degree of old-growth
associated structure into actively managed stands, while
maintaining greater timber management flexibility. This study
evaluated a 50-year growing period (no additional harvests) in
order to evaluate longer-term stand development potential
resulting from initial treatment alone. This was appropriate
because: (a) the primary research question pertained to the
restorative potential of alternate treatments, and (b) the growth
and yield models lacked adequate regeneration sub-models,
such that simulations could only project growth for the initial
population of trees. However, a 20–25-year entry cycle is
certainly a plausible option for SCE where low intensity timber
production is a management objective. This conclusion is
supported by the similar growth trends between all treatments
evaluated, including uneven-aged treatments typically scheduled on this (or a more frequent) entry cycle in northern
hardwoods.
Intermediate applications of SCE could be employed where
ecologically sensitive management is required within riparian
areas. Managing for forest structural complexity along freshwater streams would be useful where the associated influences
on in-stream aquatic habitat conditions are desired (Keeton
et al., in review). In protected areas, the full compliment of
techniques employed in SCE provide options for latesuccessional forest restoration. Managers might conduct
harvest entries once or twice, thereafter relying on accelerated
successional processes. The degree of implementation and the
number of stand entries will thus vary by application.
4.4. Silvicultural flexibility in managing stand structure
Late-successional/old-growth characteristics in northernhardwood conifer systems can be promoted through a variety of
silvicultural approaches, including those investigated in this
and other studies (e.g. Singer and Lorimer, 1997; Bryan, 2003).
Uneven-aged systems provide some (e.g. vertically complex
canopies), but not all (e.g. large live and dead trees) latesuccessional structural characteristics or provide them to a
more limited extent based on the results reported here.
Prescriptions can be modified to retain more post-harvest
structure, for instance by specifying a greater residual basal
area, larger maximum diameter, and/or lower q-factor.
However, maximum diameter limits significantly impair the
potential for large tree recruitment based on the results. Similar
findings have been reported for diameter-limit cutting (as
opposed to selection system) in uneven-aged stands (Nyland,
2005).
Cutting to alternative diameter distributions provides
another way to manipulate stand structure for a range of
objectives, including old-growth structural development. The
results show that SCE’s variable q-factor marking approach can
be used to successfully achieve a rotated sigmoid diameter
distribution. In addition, the results indicate that this distribution is sustainable in terms of growth and recruitment across
size classes. Density-dependent mortality in middle size classes
does not appear to self-thin a stand to a negative exponential
distribution, at least over the time frame investigated in this
study. This may be due to reduced (or non-constant) mortality
rates in these size classes (O’Hara, 1998; Goodburn and
Lorimer, 1999; Zenner, 2005).
Like any silvicultural prescription, use of the rotated
sigmoid will reflect stand management objectives. The
distribution would not necessarily develop in a particular
stand through natural mortality alone. Diameter distributions
in old-growth northern-hardwood forests vary considerably
(Goodburn and Lorimer, 1999). Both rotated sigmoid and
negative exponential distributions are possible, depending on
disturbance history and stand development pathway (Goff and
West, 1975; Manion and Griffin, 2001; Leak, 2002). For
silviculturists, the rotated sigmoid offers the flexibility to
manage for stand structures in which basal area distributions
are shifted towards medium and larger size classes. They
would also maintain sufficient densities of smaller stems to
ensure recruitment across all size classes. This would provide
benefits associated with larger dimension timber, large tree
habitat functions, and carbon storage associated with greater
biomass.
5. Conclusion
Foresters can manage for late-successional/old-growth
structure where desired in northern hardwood-conifer ecosystems. Active silvicultural manipulations, such as SCE, as
well as passive approaches provide alternatives for achieving
this objective. There are multiple options for retaining high
levels of post-harvest structure and for promoting accelerated
rates of stand development. These include unconventional
prescriptive diameter distributions, such as guiding curves
based on a rotated sigmoid distribution, combined with
higher levels of residual basal area, very large (or no)
maximum diameters, and/or crown release. Managing for
late-successional forest characteristics can be actively
employed as an element of structure-based forestry and
ecosystem management.
Acknowledgements
This research was supported by grants from the USDA
CSREES National Research Initiative, the Vermont Monitoring
Cooperative, the Northeastern States Research Cooperative,
and the USDA McIntire-Stennis Forest Research Program.
Nine anonymous reviewers provided helpful comments on
manuscript drafts. The author is grateful to Donald R. Tobi for
assistance throughout this project. Dr. Austin Troy assisted in
collecting and processing timber harvest volume data.
Assistance with NED software was provided by Dr. Mark
Twery. Jonathan Trigaux provided computer administration
support. Alan Howard provided statistical consulting. A special
W.S. Keeton / Forest Ecology and Management 235 (2006) 129–142
thanks goes to the outstanding field crews at the University of
Vermont.
References
Angers, V.A., Messier, C., Beaudet, M., Leduc, A., 2005. Comparing composition and structure in old-growth and harvested (selection and diameter-limit
cuts) northern hardwood stands in Quebec. For. Ecol. Manage. 217, 275–
293.
Bankowski, J., Dey, D., Boysen, E., Woods, M., Rice, J., 1996. Validation of
NE-TWIGS for tolerant hardwood stands in Ontario. Forest Research
Information Paper No. 103. Ontario Ministry of Natural Resources, Sault
Ste. Marie, Ontario.
Beatty, S.W., Stone, E.L., 1986. The variety of soil microsites created by tree
falls. Can. J. For. Res. 16, 539–548.
Bryan, R.R., 2003. Long-Term Impacts of Timber Harvesting on Stand Structure in Northern Hardwoods and Spruce-Fir Forests: Implications for
Biodiversity and Timber Production. Maine Audubon, 32 pp.
Bush, R.R., 1995. Northeastern TWIGS variant of the Forest Vegetation
Simulator. Forest Management Service Center. Available online at http:/
ftp.fs.fed.us/pub/fmsc/fvs/doc/overviews/nevar.txt.
Carlton, G.C., Bazzaz, F.A., 1998. Resource congruence and forest regeneration
following an experimental hurricane blowdown. Ecology 79, 1305–1319.
Cogbill, C.V., 2000. Vegetation of the presettlement forests of northern New
England and New York. Rhodora 102, 250–276.
Crow, T.R., Buckley, D.S., Nauertz, E.A., Zasada, J.C., 2002. Effects of
management on the composition and structure of northern hardwood forests
in upper Michigan. For. Sci. 48, 129–145.
Dahir, S.E., Lorimer, C.G., 1996. Variation in canopy gap formation among
developmental stages of northern hardwood stands. Can. J. For. Res. 26,
1875–1892.
Davis, M.B., 1996. Extent and location. In: Davis, M.B. (Ed.), Eastern OldGrowth Forests: Prospects for Rediscovery and Recovery. Island Press,
Washington, DC, pp. 18–34.
DeGaaf, R.M., Yamasaki, M., 2003. Options for managing early-successional
forest and shrubland bird habitats in the northeastern United States. For.
Ecol. Manage. 185, 179–191.
Dixon, Gary E. (comp.), 2003. Essential FVS: A User’s Guide to the Forest
Vegetation Simulator. Internal Report. USDA Forest Service, Forest Management Service Center, Fort Collins, CO. 193 pp.
Dunwiddie, P., Foster, D., Leopold, D., Leverett, R.T., 1996. Old-growth forests
of southern New England, New York, and Pennsylvania. In: Davis, M.B.
(Ed.), Eastern Old-Growth Forests: Prospects for Rediscovery and Recovery. Island Press, Washington, DC, pp. 126–143.
Elliott, C.A., 1988. A Forester’s Guide to Managing Wildife Habitats in Maine.
University of Maine Cooperative Extension, Orno, ME, 68 pp.
Foster, D.R., 1988. Disturbance history, community organization and vegetation
dynamics of the old-growth Pisgah forest, south-western New Hampshire,
U.S.A. J. Ecol. 76, 105–134.
Foster, D.R., Aber, J.D. (Eds.), 2004. Forests in Time: the Environmental
Consequences of 1,000 years of change in New England. Yale University
Press, New Haven, CT.
Franklin, J.F., Spies, T.A., Van Pelt, R., Carey, A., Thornburgh, D., Berg, D.R.,
Lindenmayer, D., Harmon, M., Keeton, W.S., Shaw, D.C., Bible, K., Chen, J.,
2002. Disturbances and the structural development of natural forest ecosystems with some implications for silviculture. For. Ecol. Manage. 155, 399–423.
Goff, F.G., West, D., 1975. Canopy-understory interaction effects on forest
population structure. For. Sci. 21, 98–108.
Goodburn, J.M., Lorimer, C.G., 1998. Cavity trees and coarse woody debris in
old-growth and managed northern hardwood forests in Wisconsin and
Michigan. Can. J. For. Res. 28, 427–438.
Goodburn, J.M., Lorimer, C.G., 1999. Population structure in old-growth and
managed northern hardwoods: an examination of the balanced diameter
distribution concept. For. Ecol. Manage. 118, 11–29.
Gore, J.A., Patterson, W.A., 1985. Mass of downed wood in northern hardwood
forests in New Hampshire: potential effects of forest management. Can. J.
For. Res. 16, 335–339.
141
Hale, C.M., Pastor, J., Rusterholtz, K.A., 1999. Comparison of structural and
compositional characteristics in old-growth and mature, managed hardwood
forests of Minnesota, U.S.A. Can. J. For. Res. 29, 1479–1489.
Harmon, M.E., Ferrell, W.K., Franklin, J.F., 1990. Effects on carbon storage
of conversion of old-growth forests to young forests. Science 247, 699–
702.
Hilt, D.E., Teck, R.M., 1989. NE-TWIGS: an individual tree growth and yield
projection system for the Northeastern United States. The Compiler 17 (2),
10–16.
Houghton, R.A., Hackler, J.L., Lawrence, K.T., 1999. The U.S. carbon budget:
contributions from land-use change. Science 285, 574–578.
Hunter, M.L., 1989. What constitutes an old-growth stand? J. For. 33–35.
Hunter, Jr, M.L. (Ed.), 1999. Maintaining Biodiversity in Forested Ecosystems.
Cambridge University Press, Cambridge, UK.
Hunter, M.L., White, A.S., 1997. Ecological thresholds and the definition of
old-growth forest stands. Nat. Areas J. 17, 292–296.
Hunter, W.C., Buechler, D.A., Canterbury, R.A., Confer, J.L., Hamel, P.B.,
2001. Conservation of disturbance dependent birds in eastern North America. Wildl. Soc. Bull. 29, 425–439.
Jenkins, J.C., Chomnacky, D.C., Heath, L.S., Birdsey, R.A., 2003. National
scale biomass estimators for United States tree species. For. Sci. 49, 12–35.
Keddy, P.A., Drummond, C.G., 1996. Ecological properties for the evaluation,
management, and restoration of temperate deciduous forest ecosystems.
Ecol. Appl. 6, 748–762.
Keeton, W.S., 2005. Managing for old-growth structure in northern hardwood
forests. In: Peterson, C.E., Maguire, D.A. (Eds.). Balancing Ecosystem
Values: Innovative Experiments for Sustainable Forestry. USDA Forest
Service, General Technical Report PNW-GTR-635. pp. 107–117.
Keeton, W.S., Troy, A.R., 2006. Balancing ecological and economic objectives
while managing for late-successional forest structure. Ecologisation of
economy as a key prerequisite for sustainable development. In: Proceedings
of the International Conference, Ukrainian National Forestry University,
L’viv, Ukraine, September 22–23, 2005.
Keeton, W.S., Kraft, C.E., Warren, D.R., in review. Structure of old-growth
riparian forests and effects on Adirondack stream systems. Ecol. Appl.
Kenefic, L.S., Nyland, R.D., 2000. Habitat diversity in uneven-aged northern
hardwood stands: a case study. USDA Forest Service Research Paper
NE-174.
King, D.I., DeGraaf, R.M., Griffin, C.R., 2001. Productivity of early-successional shrubland birds in clearcuts and groupcuts in an eastern deciduous
forest. J. Wildl. Manage. 65, 345–350.
Krankina, O.N., Harmon, M.E., 1994. The impact of intensive forest management on carbon stores in forest ecosystems. World Resour. Rev. 6, 161–177.
Krebs, C.J., 1999. Ecological Methodology. In: second ed.Addison Wesley
Longman, Inc., Menlo Park, CA, 620 pp.
Leak, W.B., 1996. Long-term structural change in uneven-aged northern hardwoods. For. Sci. 42, 160–165.
Leak, W.B. 2002. Origin of sigmoid diameter distributions. USDA Forest
Service Research Paper NE-718.
Leak, W.B., Solomon, D.S., DeBald, P.S. 1987. Silvicultural guide for northern
hardwood types in the Northeast (revised). USDA Forest Service, Research
Paper NE-603.
Liechty, H.O., Jurgensen, M.F., Mroz, G.D., Gale, M.R., 1992. Pit and mound
topography and its influence on storage of carbon, nitrogen, and organic
matter within an old-growth forest. Can. J. For. Res. 27, 1992–1997.
Lindenmayer, D.B., Franklin, J.F., 2002. Conserving Forest Biodiversity: A
Comprehensive Multiscaled Approach. Island Press, Washington, DC,
351 pp.
Litvaitis, J.A., 2001. Importance of early-successional habitats to mammals in
eastern forests. Wildl. Soc. Bull. 29, 466–473.
Lorimer, C.G., 1980. Age structure and disturbance history of a southern
Appalachian virgin forest. Ecology 61, 1169–1184.
Lorimer, C.G., 2001. Historical and ecological roles of disturbance in
eastern North American forests: 9,000 years of change. Wildl. Soc. Bull.
29, 425–439.
Lorimer, C.G., Frelich, L.E., 1994. Natural disturbance regimes in oldgrowth northern hardwoods, implications for restoration efforts. J. For.
92, 33–38.
142
W.S. Keeton / Forest Ecology and Management 235 (2006) 129–142
Lorimer, C.G., White, A.S., 2003. Scale and frequency of natural disturbances
in the northeastern U.S., implications fore early-successional forest habitats
and regional age distributions. For. Ecol. Manage. 185, 41–64.
Manion, P.D., Griffin, D.H., 2001. Large landscape scale analysis of tree death
in the Adirondack Park, New York. For. Sci. 47, 542–549.
McGee, G.G., Leopold, D.J., Nyland, R.D., 1999. Structural characteristics of
old-growth, maturing, and partially cut northern hardwood forests. Ecol.
Appl. 9, 1316–1329.
McKenny, H.C., Keeton, W.S., Donovan, T.M., 2006. Effects of structural
complexity enhancement on eastern red-backed salamander (Plethodon
cinereus) populations in northern hardwood forests. For. Ecol. Manage.
230, 186–196.
McLachlan, J.S., Foster, D.R., Menalled, F., 2000. Anthropogenic ties to latesuccessional structure and composition in four New England hemlock
stands. Ecology 81, 717–733.
Meyer, H.A., Stevenson, D.D., 1943. The structure and growth of virgin beechbirch-maple-hemlock forests in northern Pennsylvania. J. Agric Res. 67,
465–484.
Mitchell, R.J., Palik, B.J., Hunter, M.L., 2002. Natural disturbance as a guide to
silviculture. For. Ecol. Manage. 155, 315–317.
Mladenoff, D.J., Pastor, J., 1993. Sustainable forest ecosystems in the northern
hardwood and conifer forest region: concepts and management. In: Aplet,
G.H., Johnson, N., Olson, J.T., Sample, V.A. (Eds.), Defining Sustainable
Forestry. Island Press, Washington, DC, pp. 145–180.
Niese, J.N., Strong, T.F., 1992. Economic and tree diversity trade-offs in
managed northern hardwoods. Can. J. For. Res. 22, 1807–1813.
Nyland, R.D., 1998. Selection system in northern hardwoods. J. For. 96, 18–21.
Nyland, R.D., 2005. Diameter-limit cutting and silviculture: a comparison of
long-term yields and values for uneven-aged sugar-maple stands. North. J.
Appl. Forestry 22, 111–116.
O’Hara, K.L., 1998. Silviculture for structural diversity: a new look at multiaged systems. J. For. 96, 4–10.
O’Hara, K.L., 2001. The silviculture of transformation: a commentary. For.
Ecol. Manage. 151, 81–86.
Parker, G.G., Harmon, M.E., Lefsky, M.A., Chen, J., Van Pelt, R., Weiss, S.B.,
Thomas, S.C., Winner, W.E., Shaw, D.C., Franklin, J.F., 2004. Threedimensional structure of an old-growth Pseudotsuga-tsuga canopy and
its implications for radiation balance, microclimate, and gas exchange.
Ecosystems 7, 440–453.
Ripley, B.D., 1981. Spatial Statistics. John Wiley and Sons, Inc., New York, NY,
252 pp.
Seymour, R.S., White, A.S., deMaynadier, P.H., 2002. Natural disturbance
regimes in northeastern North America: evaluating silvicultural systems
using natural scales and frequencies. For. Ecol. Manage. 155, 357–367.
Shivers, B.D., Borders, B.E., 1996. Sampling Techniques for Forest Inventory.
John Wiley and Sons, New York, NY, 356 pp.
Singer, M.T., Lorimer, C.G., 1997. Crown release as a potential old-growth
restoration approach in northern hardwoods. Can. J. For. Res. 27, 1222–1232.
Sollins, P., Cline, S.P., Verhoevfn, T., Sachs, D., Spycher, G., 1987. Patterns of
log decay in old-growth Douglas-fir forest. Can. J. For. Res. 17, 1585–1595.
Strong, T.F., 1997. Harvesting intensity influences the carbon distribution in a
northern hardwood ecosystem. USDA Forest Service Research Paper NC329.
(TNC) The Nature Conservancy. 2001. Managing rich northern hardwood
forests for ecological values and timber production. The Nature Conservancy, Montpelier, VT, 18 pp.
Trombulak, S.C., 1996. The restoration of old-growth: why and how. In: Davis,
M.B. (Ed.), Eastern Old-Growth Forests: Prospects for Rediscovery and
Recovery. Island Press, Washington, DC, pp. 305–320.
Turner, D.P., Koerper, C.J., 1995. Carbon sequestration by forests of the United
States: current status and projections to the year 2040. Tellus 47B, 232–239.
Twery, M.J., Knopp, P.D., Thomasma, S.A., Rauscher, H.M., Nute, D.E., Potter,
W.D., Maier, F., Wang, J., Dass, M., Uchiyama, H., Glende, A., Hoffman,
R.E., 2005. NED-2: A decision support system for integrated forest
ecosystem management. Comput. Electron. Agric. 49, 24–43.
Tyrrell, L.E., Crow, T.R., 1994a. Dynamics of dead wood in old-growth
hemlock-hardwood forests of northern Wisconsin and northern Michigan.
Can. J. For. Res. 24, 1672–1683.
Tyrrell, L.E., Crow, T.R., 1994b. Structural characteristics of old-growth
hemlock-hardwood forests in relation to stand age. Ecology 75, 370–386.
Whitney, G.C., 1984. Fifty years of change in the arboreal vegetation of the
heart’s content, and old-growth hemlock-white pine-northern hardwood
stand. Ecology 65, 403–408.
Woods, K.D., Cogbill, C.V., 1994. Upland old-growth forests of Adirondack
Park, New York, USA. Nat. Areas J. 14, 241–257.
Zenner, E.K., 2005. Development of tree size distributions in Douglas-fir forests
under differing disturbance regimes. Ecol. Appl. 15, 701–714.
Ziegler, S.S., 2000. A comparison of structural characteristics between oldgrowth and post-fire second-growth hemlock-hardwood in Adirondack
Park, New York, U. S. A. Glob. Ecol. Biogeogr. 9, 373–389.
Ziegler, S.S., 2002. Disturbance of hemlock-dominated old-growth in northern
New York, USA. Can. J. For. Res. 32, 2106–2115.